We’re hiring a Research Scientist at IT-Universitetet i København on a project at the intersection of machine learning and operations research, funded by the European innovation community EIT Urban Mobility.
Cities across Europe are decarbonising freight - replacing diesel vans with mixed fleets of electric vehicles, cargo bikes, and light electric vehicles. This transition has enormous climate potential, but introduces optimization challenges that existing tools and benchmarks don’t capture. The goal of this project is to build an open benchmark for mixed-fleet vehicle routing grounded in real operational data from European cities, bridging the gap between academic VRP research and the reality of sustainable urban logistics.
The problem is harder than classical VRP. Real mixed-fleet operations involve vehicle-specific service times that vary dramatically by urban context (50-70% of total delivery time); vehicle-specific routing constraints where different vehicle types face different infrastructure limitations and access rules; load and elevation-dependent performance profiles that vary across vehicle types; multi-hub and multi-echelon operations with hub assignment decisions upstream of routing; and heavy-tailed uncertainty where 5% of outlier deliveries consume as much time as 70% of routine ones. Current benchmarks assume homogeneous fleets in featureless environments.
A core part of the role is working with rich operational datasets from logistics operators in Paris, Brussels, London, and Sant Cugat to understand what makes these problems hard in practice, and translating those insights into well-structured benchmark instances. The focus is not on modelling vehicle performance from scratch, but on integrating realistic operational features - context-dependent speeds, service time distributions, multi-hub structures, stochastic elements - into standardised problem formats that make it easy for the research community to develop and compare approaches. You’ll build on existing work in geospatial embeddings of urban micro-regions (H3 grids + OpenStreetMap) and vehicle-specific performance prediction.
Beyond benchmark design, you’ll evaluate and develop state-of-the-art optimization approaches against these problems, from classical metaheuristics to recent methods at the intersection of LLMs and combinatorial optimization.
This is a high-impact applied research project in close collaboration with Kale AI and European cycle logistics operators. There will be opportunities to spend time on-site with operators to understand how theory translates into daily operations and to ensure the research you produce matters beyond the lab.
You will join the DASYA lab within the DSAR section for 1 year (start date spring 2026) with the possibility of extension. The position includes funding for equipment, compute, and travel. Your basic salary, number of holidays, pension are determined by the Collective Agreement for Academics in Denmark, more information here.
Strong fit if you have:
- Background in operations research, combinatorial optimization, or applied ML
- PhD degree preferred, but not strictly required
- Experience or active interest in the intersection of LLMs and optimization
- Familiarity with open-source VRP solvers and metaheuristic frameworks
- Strong Python skills
To apply, fill in this form. Feel free to reach out by email at msia@itu.dk or nicolas@kale.ai if you have any questions about the position.
Deadline for application: 15th of March 2026.